Study highlights inaccuracy of AI chatbots in providing election information

Popular AI-powered chatbots are unreliable sources of information about the US presidential election, according to a study by the AI Democracy Projects.

 Accessories, Glasses, Paper, Pen, machine, Screw, Text, Flower, Plant, Rose

A recent study by the AI Democracy Projects, a collaboration between Proof News and the Science, Technology and Social Values Lab at the Institute for Advanced Study, has raised concerns about the accuracy of popular AI-powered chatbots in providing information about elections. The study found that when asked basic questions such as the location of polling stations or voter registration requirements, AI chatbots provided false information at least 50% of the time. The researchers tested five well-known chatbots, including Anthropic’s Claude, Google’s Gemini, Open AI’s GPT-4, Meta’s LLaMA 2, and Mistral AI’s Mixtral. The misinformation included inaccuracies about voting laws and procedures, with potential consequences for the democratic process.

In one example, the chatbots failed to accurately inform users about the legality of wearing a ‘MAGA hat’ to the polls in Texas. The study revealed that the chatbots often needed to provide more accurate and updated information about polling sites and voter registration procedures. The findings suggest a significant challenge in ensuring that AI chatbots offer users reliable and precise information, particularly in elections.

Alondra Nelson, a professor at the Institute for Advanced Study, emphasised the danger of misinformation, highlighting the study’s implications for the democratic process. The researchers expressed concerns about the cumulative effect of partially correct and misleading answers, potentially leading to frustration among voters. The study identified a need for greater caution in trusting sources for online election-related information.

Why does it matter?

The prevalence of false information provided by chatbots raises severe concerns about the information environment in which American voters are preparing for the 2024 elections. The authors noted that misinformation, including half-truths and partial truths presented as ‘artificial intelligence,’ threatens the democratic process. The study calls for increased awareness among voters about the limitations of AI chatbots and the importance of verifying information from reliable sources, especially in the critical context of elections.